Figure 2. Structural Equation Models (SEMs).

In all SEMs, PiB, flortaucipir, and GM volume were regional estimations and WMH represented a global estimation of WMH volume. We included age and APOE genotype as important modifiers of PiB, flortaucipir, WMH, and GM volume in DLB. The TIV was also added to the SEMs as a confounding factor of GM volume. (A) General form of our SEMs. (B) SEMs for regions of interest. Arrows that join nodes are direct effects, and arrows that join nodes but pass through intervening variables are indirect effects. The intervening variables are mediators. Effects come from linear regression models and represent regression coefficients (and standard errors between brackets). Coefficients were standardized per standard deviation change in each variable to facilitate interpretation across predictors. The unit for age is decades. TIV was scaled to TIV/1,000. Thickness in all arrows reflects the strength of regression coefficients, except for that of the TIV. The TIV is represented as an unweighted dashed line because this variable was not interpreted; it was included only in the SEMs to calculate the effects of PiB, flortaucipir, and WMH volume independent of TIV. Abbreviations: APOE = Apolipoprotein E; DLB = dementia with Lewy bodies; GM = gray matter; PiB = Pittsburgh compound B; SUVr = standardized uptake value ratios; WMH = white matter hyperintensities; TIV = total intracranial volume; SEM = structural equation model.